The distribution of condition numbers of rational data of bounded bit length (Q1601362)

The authors prove that rational data of bounded input length are uniformly distributed with respect to the probability distribution of the condition numbers. For the linear algebra case, it is proved that when the bit length \(h\) (the logarithm of the Northcott-Schmidt height) of a randomly choosen \(n\times n\) matrix \(M\) satisfies \(h \geq 10 n^4 \log n + \log w\), where \(w>1\), then its condition number satisfies \(k(M)<wn^{5/2}\) with probability at least \(1-2/w\). A similar estimate is established for the condition number \(\mu_{{norm}}\) introduced by M. Shub and S. Smale when is applied to systems of multivariate homogeneous polynomial equations of bounded input length. Finally, the techniques introduced are used to estimate the probability distribution of the precision, the number of bits of the denominator required to write approximate zeros of systems of multivariate polynomial equations of bounded input length.